In an effort to further test the robustness of the T Model to ferret out look-alikes based on relevant fingerprint population, 10 sample fingerprint amounts bearing 5-6 minutiae with various total values, e.g., T-Values, were searched in the County of Santa Clara AFIS database. 

Note:  The San Jose Police Department utilizes a Motorola Printrak AFIS system that applies image enhancement and minutiae detection algorithms and expert matching software to match encoded minutiae and establish candidate lists from a county of Santa Clara ten-print database of approximately 5,500,000 fingerprints [34].  If minutiae matching did not yield a definitive hit/no-hit determination, the system automatically performs expert matching. In the expert matching process the top-ranking candidates from minutiae matching are subjected to additional matching processes, including minutiae types (bifurcation or ridge endings), minutiae connectivity, ridge count, automatic ridge matching, and other data. These additional dynamic processes provide further matching accuracy and are capable of increasing the score separation between the true matching candidate and the other candidates to move the true matching candidate to the top of the list.

Different individual and combinative fingers were used as filters in order to reduce AFIS fingerprint population database size sufficiently to pre-establish the calculated number of look-alikes likely to occur for the samples searched.  Each candidate fingerprint was examined for the presence of look-alikes displaying agreement with regards to ridge formation type, ridge count, orientation and position within tolerance of friction ridge skin elasticity thresholds.  In each sample test, the model reliably identified numbers of look-alikes less than the number calculated by the model.  In no case was a look-a-like found that displayed a larger T-Value than that entered into AFIS.  

To further test the reliability of the T Model to establish insufficiency to infer positive identification based on significant amounts of corresponding ridge formations found in known non-matches, a local county of Santa Clara AFIS study was performed involving approximately 17,500 Ten-print searches, i.e. 175,000 prints, against an Unknown Latent File database containing approximately 25,000 encoded latent impressions in an effort to seek out unusually large amounts of corresponding friction ridge formations in non-matches.  The result yielded a number of non-matches containing unusually high AFIS scores.  The aggregate amounts of ridge formations in these highest scoring non-matches were examined.  The non-match with the largest and best amount of “corresponding” ridge formations, contained 11 various qualities of ridge formations with various levels of agreement quality.  All ridge formations were located in a diminishing area (see below images of AFIS Known Print and AFIS Latent Print) and based on a fingerprint population of 175,000, the number of close matches or look-alikes calculated by the T-Model clearly exceeded 1, which as a result the amount of correspondence failed to establish valid basis for sufficiency to infer identification.

AFIS Known Print 

AFIS Latent Print 

NEXT PAGE >>> 

Latent print examiners may use the T-Model Fingerprint Calculator to test and experiment with different arrangements of corresponding ridge features found in any latent v. exemplar fingerprint impressions. 

Any probabilistic model used to estimate numbers of friction ridge look-alikes must be able to reliably identify the largest and best friction ridge look-alikes that AFIS can find in its database as insufficient to infer positive identification.